當(dāng)達(dá)夢(mèng)sql monitor輸出結(jié)果太多時(shí)800行不方便查看可以考慮大力加工一下思路是創(chuàng)建一個(gè)表將結(jié)果輸入進(jìn)去然后方便用sql進(jìn)行排序和分析。et知道某個(gè)步驟最耗時(shí)但是不知道對(duì)象。大力加工創(chuàng)建一個(gè)表-- 達(dá)夢(mèng)執(zhí)行計(jì)劃采樣明細(xì)存儲(chǔ)表CREATETABLEDM_SQL_PLAN_SAMPLE(IDINTNOTNULLCOMMENT執(zhí)行計(jì)劃步驟ID,OPERATIONVARCHAR(64)NOTNULLCOMMENT操作類型CSCN2/INSERT2/SLCT2/UPDATE2等,OBJ_NAMEVARCHAR(128)COMMENT數(shù)據(jù)庫(kù)對(duì)象名表/系統(tǒng)表/視圖,ROWS_ESTIMBIGINTCOMMENT優(yōu)化器預(yù)估行數(shù),COSTBIGINTCOMMENT執(zhí)行計(jì)劃成本值,TIME_ACTIVE_SECNUMERIC(20,10)COMMENT步驟活躍耗時(shí)(秒),START_ACTIVENUMERIC(20,10)COMMENT步驟啟動(dòng)相對(duì)時(shí)間戳,EXECSINTCOMMENT當(dāng)前步驟執(zhí)行次數(shù),ROWS_ACTUALBIGINTCOMMENT實(shí)際處理/返回行數(shù),ACTIVITY_PCTNUMERIC(8,4)COMMENT活動(dòng)時(shí)間占比(%),SAMPLE_COUNTBIGINTCOMMENTActivity Detail 采樣樣本數(shù),-- 擴(kuò)展業(yè)務(wù)字段可選用于區(qū)分SQL、會(huì)話、采集時(shí)間SQL_IDVARCHAR(32)COMMENT對(duì)應(yīng)SQL唯一標(biāo)識(shí),COLLECT_TIMEDATETIMEDEFAULTSYSDATECOMMENT數(shù)據(jù)采集入庫(kù)時(shí)間,SESSION_IDINTCOMMENT會(huì)話ID);-- 建立索引按SQL_ID采集時(shí)間快速檢索CREATEINDEXIDX_DM_PLAN_SAMPLE_SQLTIMEONDM_SQL_PLAN_SAMPLE(SQL_ID,COLLECT_TIME);-- 按對(duì)象名過(guò)濾索引CREATEINDEXIDX_DM_PLAN_SAMPLE_OBJONDM_SQL_PLAN_SAMPLE(OBJ_NAME);用ai工具將這些sql monitor信息整理為insert語(yǔ)句到測(cè)試庫(kù)里執(zhí)行插入。然后就可以在manager中執(zhí)行各種排序哦原來(lái)是對(duì)SYSCOLUMNS進(jìn)行了刪除比較耗時(shí)。top對(duì)象都是數(shù)據(jù)字典表。插入動(dòng)作最耗時(shí)。對(duì)列和對(duì)象的插入刪除占大頭。數(shù)據(jù)字典是性能突破的瓶頸。